Bohan Zhuang

BLADE: Block-Sparse Attention Meets Step Distillation for Efficient Video Generation

RAPID$^3$: Tri-Level Reinforced Acceleration Policies for Diffusion Transformer

Sparsity Forcing: Reinforcing Token Sparsity of MLLMs

R-Stitch: Dynamic Trajectory Stitching for Efficient Reasoning

Motion Generalist: Multimodal Motion Generation

BlockVid: Block Diffusion for High-Fidelity and Coherent Minute-Long Video Generation

FPSAttention: Training-Aware FP8 and Sparsity Co-Design for Fast Video Diffusion

ZPressor: Bottleneck-Aware Compression for Scalable Feed-Forward 3DGS

T-Stitch: Accelerating Sampling in Pre-Trained Diffusion Models with Trajectory Stitching

Are Large Vision Language Models Good Game Players?

ZipAR: Parallel Autoregressive Image Generation through Spatial Locality

Enhancing Perception Capabilities of Multimodal LLMs with Training-Free Fusions

ME-Switch: A Memory-Efficient Expert Switching Framework for Large Language Models

Action as a Modality: Turning Multi-Modal LLMs to General Action Planners

ZipVL: Efficient Large Vision-Language Models with Dynamic Token Sparsification and KV Cache Compression